-
UNIVERSITY OF THESSALY
DEPARTMENT OF PLANNING AND REGIONALDEVELOPMENT
European Regional Development StudiesPostgraduate Program
Msc DissertationThe impact of regional specialization on economic growth: The case of
Greece
Student: Filippopoulos DimitriosSupervisor: Petrakos George
Volos, June 2012
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
2
ABSTRACT
The aim of this dissertation is to draw a basic picture of Greece’s post EU accession
experience regarding changes in the structure of manufacturing employment. For this
reason, it is undertaken an investigation of regional specialization patterns at NUTS III
spatial level disaggregated at 17 manufacturing branches according to STAKOD
classification. The dataset which is taken from ELSTAT covers the period 1980-2005
and estimations are based on the entropy index of Theil. The analysis reveals a rather
stable pattern of regional specialization. Moreover, it shows that large urban centers are
presented more diversified in relation to small-sized regions. In addition, an
econometric model is used in order to provide a possible relationship between regional
specialization and per capita Gross Value Added. The results indicate that a non-linear
relationship between the two variables has been emerged, graphically depicted by a
mirror image J-shaped pattern.
Keywords: Regional specialization, per capita GVA, Greek regions, employment,
manufacture.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
3
ACKNOWLEDGEMENTS
The completion of this dissertation would be impossible without the support and the
assistance of some people. First of all, I would like to thank my supervisor Mr. Dimitris
Kallioras for the guidance and support he showed me throughout the entire period of
study preparation. I am also truly thankful to Mrs. Maria Tsiapa whose contribution to
the completion of my dissertation was extremely important.
I would like to show my gratitude to the staff of the Department of Planning and
Regional Development, my professors and my classmates who helped me to improve
my dissertation through conversation and constructive dialogue. Finally, I would like to
thank my family and my friends for their encouragement and patience during the period
of study preparation.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
4
TABLE OF CONTENTS
ABSTRACT...................................................................................................................... 2
ACKNOWLEDGEMENTS.............................................................................................. 3
LIST OF FIGURES .......................................................................................................... 5
ACRONYMS.................................................................................................................... 6
CHAPTER 1: Introduction ............................................................................................... 7
1.1 Objectives of the dissertation.................................................................................. 71.2 Defining regional specialization ............................................................................. 81.3 Structure of the dissertation .................................................................................... 9
CHAPTER 2: Literature Review .................................................................................... 10
2.1 Regional specialization and location theories....................................................... 102.1.1 Neoclassical theory of trade........................................................................... 102.1.2 New Trade Theories (NTT) ........................................................................... 112.1.3 New Economic Geography............................................................................ 12
2.2 Specialization and Concentration: Examining their relationship through empiricalliterature ...................................................................................................................... 132.3. Specialization and Economic Growth.................................................................. 16
2.3.1 The spatial dimension of Growth theories ..................................................... 172.3.2 Structural change, specialization and growth ................................................ 182.3.3 Specialization or diversification? A policy issue........................................... 21
2.4 Empirical evidence on regional specialization ..................................................... 222.4.1 Specialization in European Union countries.................................................. 232.4.2 Specialization in European regions................................................................ 252.4.3 Econometric models ...................................................................................... 272.4.4 The case of Greek regions ............................................................................. 30
CHAPTER 3: Data and methodology............................................................................. 33
3.1 Indicators of regional specialization ..................................................................... 333.2 Description of the methodology ........................................................................... 343.3 Data presentation .................................................................................................. 36
CHAPTER 4: Specialization in Greek regions............................................................... 38
4.1 Structural characteristics of the Greek economy .................................................. 384.2 Patterns of industrial employment ........................................................................ 414.3 Analysis of specialization patterns in Greek regions........................................... 43
CHAPTER 5: An Econometric confirmation ................................................................. 49
5.1 Description of the model....................................................................................... 495.2 Interpreting the results .......................................................................................... 51
CHAPTER 6: Conclusions ............................................................................................. 54
REFERENCES ............................................................................................................... 58
APPENDIX..................................................................................................................... 63
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
5
LIST OF TABLES
Table 1: Share of industrial sector in GDP (%), 1980-2000........................................... 40
Table 2: Total employment and employment change in manufacture, 1980-2005 ........ 41
Table 3: Employment shares of industrial sectors (%), 1980-2005................................ 42
Table 4: Range of employment specialization values, 1980-2005 ................................. 45
Table 5: Regional specialization as an explanatory factor of per capita GVA (Pooled
Least Squares) at NUTSIII spatial level, 1980-2005...................................................... 51
LIST OF FIGURES
Figure 1: Comparison of Greek and EU-15 economic structures, GVA shares (%) of the
three productive sectors (primary, secondary, tertiary) for the period 1980-2005 ......... 39
Figure 2: Annual percentage growth of Greek and EU-15 industrial production (1995
constant prices), 1961-2005............................................................................................ 40
Figure 3: The non-linear relationship between specialization and per capita GVA ....... 52
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
6
ACRONYMS
ELSTAT: Hellenic Statistical Authority
EMU: Economic and Monetary Union
GDP: Gross Domestic Product
GVA: Gross Value Added
MAUP: Modifiable Area Unit Problem
NACE: Nomenclature Statistique des Activites Economiques dans la
Communaute Europeenne (in French)
NEG: New Economic Geography
NMS: New Member States
NTT: New Trade Theory
NUTS: Nomenclature of Territorial Units for Statistic
OECD: Organization for Economic Cooperation and Development
STAKOD: Statistical Classification of Branches of Economic Activity
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
7
CHAPTER 1: Introduction
1.1 Objectives of the dissertation
Internationalization process placed on a framework of globalization system have created
new conditions in worldwide transactions affecting to a great extent the productive
structure of countries and by extension specialization patterns across countries and
regions. The lowering of trade barriers, the abolishment or reduction on trade
restrictions and the remarkable progress on technological improvements in terms of
better transport and communication systems enhanced the procedure of economic
liberalization towards a more integrated economic environment (Wolfmayr-Schnitzer,
2000). The formation of this new economic environment had a remarkable impact on
government policies since each country had to be adapted to the new demands in order
to stimulate a better economic performance in its regions. In addition, the undoubtedly
dynamic presence of new economic powers such as China and India and a more
enlarged and integrated European Union which includes the ex-Soviet countries of
Eastern Europe changed the scope and the nature of global competition and therefore
played a significant role in the spatial re-distribution of economic activities. In this
framework, the study of a regional specialization constitutes a rather significant issue
which may have sensible implications on a country’s economic structure.
The current study deals with the distribution of industrial employment in the Greek
regions and the possible effect it can have on their economic performance. For this
reason, an analysis of regional specialization trends in the Greek manufacture during the
period 1980-2005 is attempted. Moreover, an econometric investigation is undertaken in
order to identify a possible relationship between specialization and per capita Gross
Value Added. The objective of this research is to find out which policy can be
considered as the most effective for a better economic potential in the Greek regions. A
policy of specialization in specific industrial sectors or a more diversified industrial
policy? In other words, in which way industrial employment should be allocated
through the regions under consideration? These are basic questions that are fully
addressed in the remainder of this study.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
8
1.2 Defining regional specialization
The observed trends in regional specialization across countries or regions are quantified
through the use of indicators. It would be therefore absolutely useful for the analysis to
provide some clear definitions about regional specialization. According to Aiginger
(1999), regional specialization is defined as the (distribution of the) shares of an
industry i in total manufacturing in a specific region r. Again, regional specialization is
the extent to which a given country specializes its activities in a relatively small number
of industries. Accordingly, a production structure of a country is said to be “highly
specialized” if a small number of industries accounts for a large share of production.
Specialization can be measured not only for production but also for exports, exports and
imports1 together and employment. On the other side, the process of a more equal
distribution of production or employment activities across industries is generally called
de-specialization or dispersion.
As it is previously referred, regional patterns of specialization are displayed through the
use of the appropriate indicators. There are several indicators used in the empirical
literature, with each presenting advantages as well as disadvantages. Whatever the case
may be, the basic distinction as regards indicators of regional specialization is between
absolute and relative measures. Absolute specialization measures the shares of
individual industries in the total manufacturing activity of a specific region.
Accordingly, a region is said to be specialized in a few industries when these industries
present high shares in the total manufacturing of this region. On the other hand, relative
specialization measures the shares of individual industries in relation to a benchmark
(the distribution of a broader geographical area). To explain this better, indexes of
relative specialization compare the distribution of industrial shares in a certain region to
the structure of a reference country. However, it is important to choose the appropriate
absolute or relative indicator in relation to the questions that should be investigated.
Thus, it is suggested by the majority of the empirical literature that absolute measures of
regional specialization should be used mainly in large countries (e.g. when we compare
1 Specialization in these cases is called “production”, “export” and “trade specialization”respectively.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
9
EU countries), whereas relative indicators should deal with the internal of countries
(e.g. when we compare regions within countries).
1.3 Structure of the dissertation
The remainder of the current dissertation is organized in five parts. Chapter two, which
follows the introduction, provides an extensive review of the theoretical framework and
the existing empirical literature with regard to regional specialization. In addition, an
analysis of the relation between regional specialization and economic growth is
undertaken in this section. Chapter three presents the dataset used in the analysis and
describes the methodological approach. Chapter four analyzes patterns of regional
specialization in Greek regions as far as manufacturing sector is concerned. Moreover,
it discusses the possible implications which can be derived from the changing patterns
of regional specialization. Chapter five examines the relationship between specialization
and per capita Gross Value Added through an econometric investigation, and finally
Chapter six summarizes the findings of the current research.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
10
CHAPTER 2: Literature Review
2.1 Regional specialization and location theories
The geographic location of economic activities and in particular concentration of
industrial activity plays a significant role in the configuration of industrial structures
especially in a status of economic integration. Therefore it is crucial to examine to what
extent traditional and contemporary trade theories can explicitly or implicitly explain
patterns of regional specialization. Recent developments in location theory try to
answer these questions by providing a wide range of evidence. In a second reading, it is
absolutely important to determine the possible impact regional patterns of
specialization could have on economic growth. Thus, theoretical elements and the
reflecting theories which explain changes in regional specialization and geographic
concentration must be carefully examined. However it must be highlighted that none of
these theories and hypotheses alone has been proved sufficient to fully explain the
determinants of industrial location. A brief summary of location theories is presented
below in order to be conceived the main determinants of the interaction between space
and industrial activities.
2.1.1 Neoclassical theory of trade
International trade theory has severe impacts on regional specialization and industrial
concentration patterns and as Isard (1956) pointed out spatial location of economic
activity and trade are the two sides of the same coin. Neoclassical theory has fairly
characterized by Krugman (1993) as “first nature”, paying particular attention to natural
(factor) endowments and technology for determining the spatial dimension of economic
activity. The neo-classical trade theory – assuming perfect competition, constant returns
to scale in production and a market with homogeneous products as the determining
factors in these models – has tried to explain regional specialization through the notion
of comparative advantage in terms of the availability of natural recourses and
technological level. Ricardo’s (1817) “comparative advantage” refers to cross-country
differences in the productivity of labour as the only factor which can explain
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
11
differences in comparative production costs. On the other hand Heckscher-Ohlin theory
of trade focuses mainly on factor endowments assuming that technology is similar
across countries. Therefore, differences in production can be explained by differences
in factor endowments or differences in the abundance of production factors [Heckscher
(1919), Ohlin (1933)].
2.1.2 New Trade Theories (NTT)
On the other hand new models of trade theories – assuming imperfect competition,
increasing returns to scale and differentiated products – have emerged to point out that
comparative advantage could not be considered to be the only sufficient explanation for
regional specialization due to the fact that regions and particularly countries do exhibit
completely different production structures. New trade theories have been developed in
an attempt to supplement the traditional neoclassical trade theory explaining the notion
of intra-industry2 trade as the main determinant in a framework of monopolistic
competition and differentiated products. However, this does not mean that New Trade
Theories exclude the existence of inter-industry trade among countries as both intra and
inter-industry forms of trade take place to the theoretical framework of New Trade
Theories. In this procedure the most important element in the theoretical modeling of
NTT is the role of market access. The latter can be explained by the industrial
concentration in countries that exhibit good access to large markets. Assuming
immobility of production factors, firms tend to concentrate in large markets where
industries can exploit scale economies and take advantage of lower trade costs due to
the large domestic demand. Krugman (1980) made this clearer by what has become
known as the “home market effect”. The explanation for “home market effect” stems
from the ascertainment that, ceteris paribus, countries tend to export those goods for
which they have relatively large domestic markets. Consequently, in a model of two
countries, each country specializes in types of products for which it has the larger home
market and thus it becomes a net exporter of these products.
2 Intra-industry trade is characterized by an exchange of differentiated goods which belong inthe same product category (same industries)
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
12
2.1.3 New Economic Geography
The New Economic Geography that has emerged recently maintains the basic
assumptions made by New Trade Theory namely monopolistic competition and
increasing returns to scale. The new distinctive characteristic of NEG is Krugman’s
(1991a, 1991b) assumption that labor is an internationally mobile production factor. In
this framework he shows that due to the interaction between scale economies, trade
costs and international mobility of labor, two initially identical countries may give rise
to an industrial core and a periphery. Thus, agglomeration of economic activities forces
industrial firms to locate in regions with larger market share because they can better
exploit economies of scale taking advantage of an extensive labor force and sharing
specialized input suppliers. A second class of NEG models proposed by Venables
(1996) assumes that labor is internationally immobile but allows for input-output
linkages between firms. To put it simply, producers of final goods (downstream firms)
seek to locate in a market comprised of many upstream firms3 lowering in such a way
transport costs. The demand and cost linkages or else backward and forward linkages
created by vertically related firms represent the driving force that can trigger
agglomeration. In these models a reduction in transport costs can lead to increased
specialization and concentration but at very low levels of transport costs dispersion
trends are likely to appear. To sum up, scale economies, spillovers and forward and
backward linkages function as centripetal forces whilst costs incurred by agglomeration
such as commuting and congestion costs function as centrifugal forces (Fujita,
Krugman, Venables 1999). Conclusively, at intermediate trade costs industries prefer to
concentrate at the core taking advantage of a larger market even if wages are higher in
relation to the periphery, while industries tend to move to the periphery in order to be
benefited from lower wages at very low levels of trade costs. Whatever the case may
be, these models follow specific assumptions and function under particular
circumstances. We must therefore be very cautious when we try to interpret the
operation of these models to reality. The fact is that each model alone can explain a part
of reality but in any case they cannot explain the whole truth.
3 Upstream firms are the producers of intermediate goods whereas downstream firms are theproducers of final goods
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
13
2.2 Specialization and Concentration: Examining their relationshipthrough empirical literature
Globalization and trade liberalization have induced dramatic changes in global
production and consumption and this unequivocally does affect the productive
structures of countries especially when they are in a status of economic integration.
European Union constitutes a special example of economic integration having created a
single market and a single currency in part. This procedure has produced severe
implications in national and regional level affecting to a great extent the structure of
European manufacturing and afterwards patterns of regional specialization and
industrial concentration. In this respect, another crucial question that literature of spatial
economics has examined is the relationship between regional specialization and
geographic concentration. Accordingly, are there specific characteristics between
countries and industries that could explain the differences in specialization and
concentration patterns? What are the driving forces which determine the location choice
of industries and which factors drive them to change their behavior over time?
Although traditional trade theory, new trade theory and new economic geography bring
into light some useful insights about this possible relationship they do not provide clear
and definite predictions about this relationship. As Aiginger and Pfaffermayr (2004)
point out “some determinants are addressed in trade theory, some in industrial
organization and some in economic geography”.
It is therefore crucial to examine thoroughly the consistency of predictions made by
traditional and contemporary location theories with industry characteristics basically in
the light of EU experience. Economic integration within the European Union dropped
the trade barriers in favor of further trade liberalization allowing for free movement of
goods and people. Thus, in addition to theoretical models of traditional trade theories
which are based on comparative advantage and factor endowments, new trade theories
draw attention to the role of market access and the interaction between scale economies
and trade costs concerning both the characteristics of the industries and the
characteristics of the countries where industries locate (Amiti 1998). Starting from
traditional trade theories, one could say intuitively that specialization according to
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
14
comparative advantage affects significantly the pattern of relative concentration4 while
says nothing about absolute concentration. However, in the case of New Economic
Geography which emphasizes in industrial agglomerations stimulated by forward and
backward linkages between firms, the appropriate measure should be the absolute
concentration (Haaland et al. 1999). In this sense, one can conclude that specialization
according to comparative advantage fits well to small labour-based countries while
specialization explained by home market effect and agglomeration forces has to do with
larger and more central – as market access considered – countries. Indeed, Haaland et
al. (1999) find that industries like Motor Vehicles, Electrical Apparatus, Machinery and
Equipment, Radio, TV and Communication Equipment “are among the most
concentrated ones in terms of absolute concentration, whereas there are not
particularly concentrated in relative terms”. This is the case of industries that can
exploit high levels of scale economies implying that are basically concentrated in large
countries. On the other hand, industries like Railroad Equipment, Wearing Apparel and
Shipbuilding and Repairing “are fairly concentrated in relative terms, but not in
absolute terms”. The latter indicates that small countries are mainly specialized in this
type of industries. Brulhart (1998) comes to confirm the above observations regarding
country specialization in light of concentration of industrial sectors. From the
estimation of locational Gini index between 1980 and 1990, he finds a considerable
increase of industrial concentration in 14 out of 18 sectors with respect to
manufacturing employment. There is also evidence that industries subject to high scale
economies are highly concentrated and located in central EU countries. But the most
interesting point in his analysis is to see in which way specialization patterns of
individual countries reflect the increasing trend in concentration. The following
example shows the general tendency. On the one side Portugal which is regarded as a
peripheral country presented in 1990 the highest level of specialization in labor-
intensive sectors such as Textiles and Clothing/footwear, while the Netherlands
exhibited the lowest value in these sectors. On the other side, Germany – which belongs
to the strong European core –, appeared to be the most specialized country in Motor
Vehicles and Electrical Engineering while the opposite is true for Greece. In terms of
overall manufacturing employment the same stylized fact is applied: Germany is the
4 Relative concentration measures to which degree an industry is concentrated relative to theaverage spread of activities between countries, while absolute concentration indicates whetheran industry is concentrated in absolute terms (Haaland et al. 1999)
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
15
most specialized member, whereas Greece – the EU’s most peripheral country – is the
least specialized. It therefore becomes tangible that peripheral countries are specialized
in low-scaled and labor-intensive activities, whilst more central countries concentrate
high-scaled, high-technology and capital-intensive activities.
In addition, there is also another element that should be taken into consideration in the
examination of regional patterns of specialization and industrial concentration: the
possible connection among them. Are regional specialization and geographic
concentration the two sides of the same coin? In other words do the two concepts move
in the same direction as regards industrial structures of countries or regions? One might
suppose that a country or region which becomes more specialized in a few industrial
sectors, it probably concentrates more of its activity in these sectors. But in a world of
asymmetries, different population sizes and differences in factor endowments and
technology it is not that simple. Aiginger and Davies (2004) using production data in
their analysis suggest that although specialization of European manufacturing has
showed an increasing trend, concentration has moved in the opposite direction with
respect to the period 1985-1998. The results form a different picture if we analyze the
data for the two sub-periods, 1985-1992 and 1992-1998. Between the period 1985-1982
which is defined as the Pre-Single Market period industries became more concentrated,
while in the second sub-period a decrease in geographical concentration had been
observed. This view is also supported by Aiginger and Rossi-Hansberg (2006). They
used two data sets on manufacturing activity across the United States and the European
Union member states for the period 1987-1996 and showed that for a broad set of
transport costs specialization increases and concentration decreases as transport costs
fall. With respect to specialization Amiti (1997) finds that “even though specialization
decreased for some countries when comparing 1968 and 1990, there was a significant
increase in specialization between 1980 and 1990 in all of them”. It can therefore be
implied that the impact of the Single Market implementation in the European Union is
undoubtedly of particular significance. The trends of industrial de-concentration during
the Single Market period at the early nineties have also been confirmed by Aiginger
and Pfaffermayr (2004) either by using value added or employment or even export data.
As for the Pre-Single Market period and especially during the 1980s Brulhart (1998),
Brulhart and Torstensson (1996), Amiti (1998) and Haaland et al. (1999) also provide
evidence of increasing trends in geographical concentration. Haaland et al. (1999) find
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
16
that on average relative concentration increased by 11.4 % during the period 1985-1992
and only few industries exhibited decreased concentration. The main conclusion is that
concentration and specialization went together until the early 1990s but from this point
onwards they did not develop in parallel. All in all, the empirical research has
confirmed the stylized fact that the enactment of the Single Market during the 1990s
leaded to a significant decrease in overall geographical concentration in the EU
territory.
2.3. Specialization and Economic Growth
International trade theories have shown that the nature of the specialization of a country
is non-neutral on its growth performance. However most empirical studies related to
growth literature do not take into account the potential effects of specialization on
growth (Bensidoun et al. 2001). In addition, it is observable a lack of research in this
field – connection between specialization and growth – and thus further observation is
required in order to be determined a possible relation among the two. Empirical
literature must therefore seek to answer in the following questions:
-Do the specific types of industries which countries are specialized in provide evidence
of a more growth motivating economy?
-Does the industrial sector composition across countries or regions constitute a major
factor of explaining growth rates?
-In other words, what is the best strategy that promotes growth in a country as far as
manufacture is concerned? Regional specialization or regional diversification?
The answer in the latter is not so obvious due to the fact that several features –
endogenous or exogenous in nature – that induce growth should be taken into
consideration before a clear policy of specialization or diversification is adopted.
Furthermore, the choice of the appropriate strategy constitutes an issue of high
importance regarding its impact on personal income, employment, value added, the
level of education and other determining factors of economic growth.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
17
2.3.1 The spatial dimension of Growth theories
Economic growth and its determinants have attracted the attention of theoretical and
empirical literature especially over the last decades. As far as growth theories are
concerned, it is worth mentioning that “due to the lack of a unifying theory on
economic growth […] studies draw on several theoretical frameworks and examine
factors that are taken from several sources” (Arvanitidis et al. 2007). It is therefore
easily understood that findings and conclusions of these studies are at least insecure and
often contradictory. However, despite the lack of a unifying growth theory, there are
several theories that can partially explain the role of growth determinants and their
impact on regional income. At this point it is essential for the purpose of the analysis to
examine which of these theories can include in their framework the component of
spatial dimension. The conventional neoclassical model of Solow (1956) which
assumes constant returns to scale, substitutability between labor and capital and an
exogenously determined technological progress, it does not provide signs of how
industrial activity can be distributed in space. The model shows how the interrelation
between the increase in accumulation of capital, the increase in workforce and
technological progress can affect the aggregate income of an economy. However,
despite the fact that technological progress is regarded as a major factor in this model,
its exogenous nature does not allow for any spatial interpretation. On the other side
endogenous growth theories (Romer 1986; Lucas 1988)5 operating in a context of
increasing returns to scale, highlight the role of factors such as the accumulation of
knowledge and innovation. The introduction of these factors in these models aims to
endogenize the process of technological progress causing in such a way a self-powered
economic growth. Whatever the case may be, it seems that endogenous growth models
are likely to play an important role as regards spatial dimension. Due to the fact that
endogenous theories leave room for state intervention in the forms of national and
regional policies the above statement may intuitively be true. Another strand of theory
which moves in the same direction with the previous is the cumulative causation
growth theory (Myrdal 1957; Kaldor 1970). The basic point of this theory is that
economic activity is not evenly distributed across space and that “initial conditions”
5 Romer’s (1986) model explains growth through technological externalities such as learning bydoing and knowledge spillovers, while the basic role in Lucas’ (1988) model plays humancapital.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
18
play a decisive role in the determination of economic growth. This growth process
generates unbalanced regional growth as powerful regions reinforce their position
increasing the distance from the weak regions. Cumulative causation theory seems to
present some striking similarities with the New Economic Geography (Krugman
1991a) although NEG is not regarded a growth theory. Despite the fact that NEG has to
do with location of economic activity, it also has severe implications on economic
growth.
2.3.2 Structural change, specialization and growth
The presence of income differences across countries but even across regions has given
rise to a continuous empirical research in order to identify possible factors that induce
growth. In this respect it is of high importance the examination of the impact that
sectoral composition of economic activity can have on regional growth. This
phenomenon has been mainly explored in European Union where extensive structural
change has taken place in the light of economic integration. However, while most work
try to explain growth differences by focusing on structural characteristics and other
variables such as human capital and level of technology, few studies use specialization
as a determining factor of growth rates in a country. As Aiginger (2001) rightly argues
“the relation between structural change and growth seems to be under-researched
relative to its alleged importance” since very few studies consider the interrelation
among the two. The impact that structural change could have on economic dynamics of
a country or even region must be therefore faced with particular attention from the
scientific community.
Most empirical research has focused so far on the examination of specialization of
countries and concentration of industries leaving unsearchable the possible relation
between specialization and economic growth. However, there are studies that have
attempted to analyze how changes in spatial allocation of industrial activity can affect
the economic potential of countries implicitly or explicitly. Peneder (2002) referring to
the connection between structural change and aggregate growth suggests the
confirmation of three general lessons:
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
19
-Firstly, industries generally do not contribute equally to overall growth in labor
productivity.
-Secondly, structural change itself is not a uniform process since it is more pronounced
for some industries in certain periods, and less in others.
-Thirdly, there is a tendency for structural change during periods of low aggregate
growth.
The suggestions made by Peneder underline the fact that it is very difficult to define a
clear and monotonic relation between observed structural change and aggregate growth
as there is evident an uneven distribution of industrial activity across space and time.
Moreover, it is also difficult to determine a direct one-way causality – whether growth
depends on past change or whether growth promotes structural change – as regards the
two variables. In the same line Aiginger (2001) argues that growth provokes structural
change, but on the other hand a change in industrial structures is a precondition for
growth. He nevertheless finds evidence that growth depends on past structural change
more closely than the other way round. Using nominal and real value added, and
employment as variables in his study, Aiginger finds support for a close relation
between speed of change and growth of manufacturing regarding European Union for
the period 1985-1998. The only exception which reduces the closeness of the fit is
Greece in which structural change is considerable while growth is appeared to be the
lowest in the EU.
Another stylized fact presented in both studies (Aiginger 2001; Peneder 2002) is the
positive relation between the levels of economic development and specific kind of
industrial structure. Peneder (2002) finds that within the manufacturing sector both
technology driven and high skill industries present a significant and positive impact on
the level of GDP per capita, confirming the fact that fast growing industries can achieve
higher rates of productivity growth than others. Aiginger (2001) moves in the same
direction stressing that increases in the shares of fast growing industries6 and decreases
in opposite kinds of industries are considered to be growth promoting for a specific
country. According to economic theory, rising incomes induce changes in demand
6 Aiginger (2001) entitles this kind of positive changes “active” change, while he refers to“passive change” as far as negative changes -increases in slowly growing industries anddecreases in fast growing industries- take place.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
20
structures and thus countries must specialize in growth promoting sectors adapting their
production structures in accordance with changing demand structures. This means that
countries or regions must proceed a systematic re-shaping of specialization patterns in
order to be adjusted to the new demand requirements. An important interpretation can
be implied from the latter statement: changes in specialization patterns induced by
structural changes may implicitly affect economic growth if not explicitly.
The adjustment process to new market conditions could be the case for the countries
which belong to diverging clubs7 or “the poor countries”. These countries have to
follow another specialization strategy provided that they need to succeed better growth
rates. Bensidoun et al. (2001) explain that these countries have presented better
catching-up performance when they succeed to adapt their international specialization
to dynamic products or else in products that incorporate a dynamic international
demand. This fact is also confirmed by Bensidoun and Ünal-Kesenci (1998) and
Grossman and Helpman (1991) who point out that specialization in high-technology
and high-quality sectors and generally in increasing returns sectors can only provide
better results as regards growth performance. On the other side, countries that do not
follow this strategy and insist on traditional production structures are characterized by
low share in world trade and thereupon by poor growth performance. The latter seems
to be the case for the regions of European Union. In a study of European Union regions
during the period 1977-1999, Ezcurra et al. (2004) find that changes in regional
specialization patterns are closely linked to the distribution of regional GDP per capita.
They suggest that the increase in regional specialization during the nineties may explain
the presence of regional inequality and the maintenance in the degree of polarization of
regional per capita income. It can be therefore implied from this that specialization of
low-income countries in sectors of low growth potential has negative effects on their
economies especially in a status of economic integration.
Furthermore, a basic point that must be explored through the scanning of scientific
literature is the possible role specialization may have on growth determinants such as
productivity and employment. There are several studies that confirm this relation whilst
others do not find an explicit relation between the two. Weinhold and Rauch (1997)
7 For an overview about converging and diverging clubs see Quah (1996)
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
21
suggest that in a state of openness – where economies can take advantage of dynamic
scale economies associated with learning by doing – regional specialization can have a
positive impact on productivity growth. However, Combes (2000) looking at the
economic structure and local growth for 341 French spatial entities over the period
1984-1993 finds evidence that regional specialization negatively affects employment
growth. He, nevertheless, stresses the fact that specialization may improve local growth
in expansion periods while the opposite is true during recession periods.
2.3.3 Specialization or diversification? A policy issue
At this point, another crucial aspect of economic growth that must be examined is the
choice of the appropriate strategy between specialization and diversification. Do
specialization or diversification trends across regions or countries coincide with
increases in per capita incomes or declines? As Aiginger (1999) points out “no
comprehensive empirical investigation is available on the topic whether higher
specialized countries or those with a more dispersed structures - across industries or
locations - are better for growth”.
As it has been suggested from many studies, specialization in specific growth-
promoting sectors such as high-technology or more generally scale-intensive industries
can evidently foster economic growth. But can regional specialization be proved an
effective policy which can be applied to countries without putting them in a state of
jeopardy? Dalum et al. (1999) stresses that specialization in the “right” kind of
activities may be successful but he also suggests that “enhancing growth by steering
specialization patterns seems a quite risky art rather than a well-established science
without major uncertainty”. Aiginger (1999) and Ezcurra et al. (2004) referring to the
EU case point out that specialization in narrow product groups may increase demand
risk for individual countries and this possibly will make them more vulnerable to
asymmetric shocks especially when these countries belong to a common currency area.
It is obvious that external shocks – especially for the countries of a Monetary Union –
can lead to severe demand asymmetries which cannot be faced by changes in the
external value of currencies. On the other hand, countries which present a more
diversified industrial structure will be in a more advantageous position than others
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
22
(Aiginger 2001). However, Bode et al. (2004) examining sectoral specialization and
performance of the Spanish regions showed that diversification coincides with slow
growth, while specialization with quick growth. This does not seem to be the case for
the peripheral regions of European Union over the period 1950-1990. Molle (1997)
finds out that the lower levels of GDP per capita have been presented in those
peripheral regions which exhibited higher levels of specialization.
Whatever the case may be, it is beyond any question that specialization not only
presents advantages with regard to growth potentials, but also performs major
disadvantages related to risk effects. Specialization in dynamic markets give countries
the chance to enjoy higher levels of productivity and accordingly higher economic
growth, while countries specialized in mature, low-wage or low-growth potential
industries will not be able to achieve faster growth (Aiginger 2001).
2.4 Empirical evidence on regional specialization
A considerable number of empirical studies related to the estimation of specialization
across countries, regions or more generally geographical entities have been exhibited
over the last years especially in the European context. However, there is an observable
lack of information in this field, since most of studies deal with specialization in
European countries and empirical evidence at the level of European regions is
particularly sparse (Krieger-Boden 2000). Whatever the case may be, the thorough
examination of regional specialization has been proved to be a very effective tool for
policy makers due to its particular importance in both economic and political terms.
The main focus of this review will be the exploration of regional specialization trends
in the European Union which forms a geographical location of high interest due to its
distinctive spatial specificities. The extensive European integration that took place over
the last decades has nevertheless showed that the mobility of labor appears to be rather
limited with respect to EU-15 (Fertig and Schmidt 2002; Fertig 2003), hence only
marginal changes in the degree of specialization of member states have occurred
(European Commission 1999a).
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
23
Most empirical research in the field of regional specialization refers to the
manufacturing sector due to the availability of data sources. Trade, production and
employment data are used in this direction in order to be examined the role of
specialization in both higher (e.g. countries) and lower (e.g. regions) level of spatial
aggregation.
2.4.1 Specialization in European Union countries
Various studies that deal with European countries concentrate their analysis in a basic
question: Have economic integration affected patterns of regional specialization over
the last years? In other words do EU member states present increasing or decreasing
trends of specialization?
Firstly, Hine (1990) and Greenaway and Hine (1991) find evidence of increasing
specialization as regards EU countries in the early 1980s. The results of their survey are
based on the estimation of the mean of the Finger-Kreinin index (F-K), using
production and export data for 28 manufacturing industries. On the contrary, Sapir
(1996) comes to a different conclusion regarding specialization in EU countries. His
analysis is based on the estimation of Herfindahl index with trade data from 100
manufacturing industries. He finds that specialization did not changed in Germany,
Italy and the UK for the period 1977-1992, while increased in France since 1986.
A comprehensive analysis of specialization trends in EU member states was conducted
by Amiti (1997). She uses two databases – one from Eurostat and the other from Unido
– and considers the estimation of two measures of specialization, the Gini (Gj) index
and the weighted standard deviation of the Balassa index (sj) using production and
employment data. The Eurostat dataset includes 65 manufacturing industries and
presents results for five European countries (Belgium, France, Germany, Italy and the
UK), while the Unido dataset consists of 27 manufacturing industries and 10 European
countries namely Belgium, France, Germany, Italy, Denmark, Greece, Portugal, Spain,
the Netherlands and the UK. In the case of Eurostat dataset she finds increasing
specialization at an average annual rate of 2% in all countries for the period 1976-1989.
In the second case of Unido dataset the results are mixed but the general trend is
increasing. More specifically, between 1968 and 1990 there was a significant increase
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
24
in specialization for Belgium, Denmark, Germany, Greece, Italy and the Netherlands, a
significant fall for France, Spain and the UK, and no significant change for Portugal.
However, it is remarkable that France, Spain, Portugal and the UK exhibited upward
trends for the period 1980-1990. Amiti (1997) argues that the latter is possibly the
outcome of the elimination of trade barriers within the EU especially for countries that
are late joiners to the EU.
Almost the same results are applied to Midelfart-Knarvik et al. (2000) survey. They use
as the main data source the OECD STAN database for 14 European countries (the EU-
15 except Luxemburg) and estimate Krugman specialization index using production
data over the period 1970-1997. Although a fall in specialization is observable between
1970-1980, there is evident a steady increase from 1980 onwards in all countries except
the Netherlands. This consequently leads to the conclusion that from the early 1980s
industrial structure of each individual country tended to be more dissimilar in relation
to the rest of the EU. Midelfart-Knarvik et al. (2000) draw attention to this feature and
estimate the bilateral differences between the industrial structures of pairs of countries.
The basic point which can be excluded from this comparison is that countries of
European core (e.g. Germany, France, GB) appear to be more similar each other and
the same is true for peripheral countries (e.g. Greece, Portugal). However, when the
first group is compared to the second, there is evident an increasing degree of
dissimilarity, confirming in such a way an established core-periphery pattern. The
steady increase in specialization of EU member states from 1980 onwards is also
evident in Aiginger and Davies (2004). Having used nominal value added data for 14
countries (Belgium and Luxemburg are taken together) and 99 manufacturing
industries, they estimate the entropy index8 of specialization and find that countries
became more specialized during the period 1985-1998. Besides, the main point of their
analysis is that specialization grew faster during the nineties after the full introduction
of the Single Market, having presented a change of 5% in a period of 6 years (1992-
1998). The above consideration is also confirmed by Aiginger and Rossi-Hansberg
(2006) who come to the conclusion that average specialization in European Union
countries rose by 5.7% for the period 1987-1996. Furthermore, they go through a
8 The used entropy index SPEC j = - jiji jij XXXX /ln*/ is defined by the summationof the products of the shares and log shares of each industry in the country’s aggregatemanufacturing (Aiginger and Davies 2004).
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
25
comparison between United States and European Union specialization degrees and
conclude that average specialization grew faster in the EU (EU 5.7% ; US 2.3%) for the
same period. For their survey, they use Gini coefficient as the appropriate index for two
datasets; one for 50 US countries and 10 industries and the other for 14 EU countries
and 23 industries.
On the other side, no clear tendency towards increasing or decreasing specialization for
the period 1980-1994 has been detected by Krieger-Boden (2000). The estimations of
coefficients of specialization9 for value added and employment for 12 EU countries
leads to ambiguous results, since some countries show a slight increase while others do
not present any clear trend. The survey of the European Commission (1999b) comes to
the same conclusion as there is no general trend of increasing specialization10 or
increasing diversification over the period 1988-1998. However, it is evident that
although production specialization exhibits increasing trends in the majority of member
states, export specialization presents a downward trend in almost all countries.
2.4.2 Specialization in European regions
Until recently, most empirical studies related to specialization in European Union have
used national data (e.g. data at country level) and not regional. The lack of empirical
results at a lower territorial level was mainly due to a severe lack of data on European
regions. As it can be observed from the literature, the time periods that have been taken
in most surveys are extremely short by virtue of insufficient industrial disaggregation
found in most European regions. Using GVA data from Eurostat REGIO database,
Hallet (2000) tries to find out trends in sectoral specialization11 for 119 European
regions. For this purpose, he estimates the absolute difference between the sectoral
share y ki of branch k in region i and the respective EU15 averageky , summed over all
9 s = n
iii ba , where ia are the industrial shares of the country under investigation and ib
are the industrial shares of a reference economy (e.g. EU average), where 0≤s≤2 (Krieger-Boden, 2000).10 The results are based on the estimates of 7 indicators of specialization for 14 countries andtwo levels of aggregation.11 He actually uses the sectoral classification NACE 17, which comprises 17 branches ofeconomic activity and includes 5 groups of services.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
26
branches k12. He finds that between 1980 and 1995 regional specialization presented a
decreasing trend, as only 34 out of 119 European regions have become more
specialized. However, a clear pattern of specialization cannot be identified by the
results because the regions that became more specialized during this period are either
among the poorer regions or among the richer ones. The study of Hallet comes to
confirm the results of a similar study conducted by Molle (1997) who finds a general
decreasing trend in specialization for a longer time period, 1950-1990. With respect to
within countries analysis, Bode et al. (2004) examines the evolution of regional
specialization in Spain with the use of Theil index and Weighted Theil index.
Employment data disaggregated into 18 Spanish regions and 88 manufacturing
branches reveals that during the period 1978-1999 specialization of Spanish regions
seems to have been moderate. Furthermore no clear tendency of increasing or
decreasing trend in regional specialization has been observed for this period. Having
used employment data Krieger-Boden (2000) examines regional specialization in
France for the period 1973-1996. Herfindahl and Gini indices have been calculated for
21 regions and 30 manufacturing branches, but the outcome seems to be rather
contradictory. According to the results, Herfindahl index reveals no variation as regards
specialization, whereas the estimation of Gini coefficient shows that specialization in
most regions has presented decreasing trends.
During the last decade, European Union carried out the greater enlargement in its
history, accepting countries of former Eastern bloc as new member states.
Specialization patterns in the regions of these countries especially from 1990 onwards,
when they start functioning in a state of free market, have been extensively explored by
the empirical literature. Traistaru et al. (2002) analyze trends in specialization patterns
during the period 1990-1999 for the accession countries of Bulgaria, Romania, Estonia,
Hungary and Slovenia using regional manufacturing employment data at NUTS III
spatial level. They find that average regional specialization13 increased in Bulgaria and
12 The equation is formed as follows: s i = k
kki yy2
1
13 They use as a measure of regional specialization the Dissimilarity Index:DSR j = i isij ss , where s sij is the share of employment in industry i in region j in totalemployment of the region and s i is the share of country employment in industry i in totalcountry employment.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
27
Romania, decreased in Estonia and it did not exhibited any significant change in
Hungary and Slovenia. Also it can be observed from the analysis that highly-
specialized regions reveal higher GDP per capita than low-specialized regions. For the
same group of countries and the same time period, Kallioras et al. (2004) – with the use
of Theil entropy index estimated for NUTS III regions – find that countries with
intermediate economic level such as Hungary and Estonia presented prominent changes
in the degree of regional specialization, whereas countries with high (Slovenia) or low
(Bulgaria, Romania) level of economic development were characterized by stable
industrial patterns. In addition, Kallioras (2006) points out that during the period 1990-
2000 the majority of regions in EU accession countries recorded a general decreasing
trend in the degree of specialization as measured by Theil index. However in some
cases, – mostly for the regions of Hungary, Estonia and Slovenia – regional
specialization exhibited increasing trends mainly due to the durability of productive
bases of the respective regions. In this framework, it is of high interest the observation
made by Resmini (2002) who stresses that relocation activity of manufacturing sector
was very intensive during that period and mainly in favor of regions which border the
EU. As a result, specialization levels in most border regions – but also in capital cities –
presented upward trends and better growth levels as compared to the rest of the regions.
The latter comes to confirm the crucial role European integration process has played to
the structure of industrial sector in EU accession countries.
2.4.3 Econometric models
Theoretically, it is admissible by the literature that regional specialization can influence
the growth prospects of countries and regions. However, the impact regional
specialization can have on per capita income has not been explicitly proved by the
scientific research. At the same time, spatial econometric analysis has revealed in some
cases that specialization – especially in industrial sector – matters for growth. Indeed,
changes in regional specialization together with other determinants of growth such as
regional population, density of population, investments or technology appear to be
depicted by changes in per capita income. The majority of econometric models use
regional specialization as independent variable, while regional per capita GDP is
applied in most cases as the dependent variable in the models under consideration.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
28
Below, a further investigation of spatial econometric models is attempted in an effort to
be understood the interplay between growth and specialization.
Bensidoun et al. (2001) in a study of 53 countries for six periods of 5 years (1967-1997)
examine the interrelation between international specialization and growth with the use
of a dynamic panel-data model. The general form of the equation is the following:
ln ity -ln ity =α i +βln ity +δ1 ln itinv +δ 2 ln itdisc +λln itspec + t + it ,
where ity is the PPP14 GDP per capita of country i at time t, itinv is the investment rate
for the period from 1-τ to t-1, itdisc is an indicator of openness and itspec is the
specialization indicator. From the estimates it can be concluded that the nature of
specialization or more specifically the ability of countries to adapt to new demand
conditions relates positively and significantly to growth. According to the authors,
specialization in dynamic products may be proved growth promoting, since
specialization in specific products is better for growth than specialization in other less
dynamic products.
Dalum et al. (1999) stresses the importance of specialization on economic growth
through a study of 20 OECD countries15 for the period 1965-1988. They use export data
for 75 industrial products, each of which belongs to one of 11 manufacturing sectors
and estimate separate equations for three periods16: 1965-1973, 1973-1979 and 1979-
1988. The model used for this analysis can be written as:
ijtQ =α jt L ijt +β jt K+γ jt ijtT + jt ijtU + jts ijtS
where Q is value added, L is labor input, K is capital input, T depicts technology
investment, U is a proxy for international technology diffusion and finally S is a vector
of specialization variables. The regression results indicate that specialization does
14 PPP or Purchasing Power Parity is an alternative measure of GDP15 Austria, Belgium, Canada, The Netherlands, Portugal, Spain, France, Germany (West),Switzerland, Denmark, Sweden, Norway, Finland, Japan, the United Kingdom, the UnitedStates, Greece, Turkey, Ireland, Italy16 The period of analysis is divided into three sub-periods because the authors try to catch thecyclical variations in export and exchange rates. The years 1965, 1973, 1979 and 1988 areregarded as peaks in the business and trade cycles.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
29
matter for growth, even if the effect becomes less important over time. It seems that
specialization in combination with other factors such as technology and knowledge
spillovers can evidently explain growth, despite the fact that more work in this field is
essential.
With respect to European Union, Ezcurra et al. (2004) present an econometric model in
which regional productive specialization is considered to be the dependent variable,
while regional per capita income plays the role of the explanatory variable. The model
is as follows:
KitSPEC = 0 + 1 log itPOP + 2 log itDENS + 3 logGV itApc + 4 log 2log itGVApc +
5 iCENTRAL + 6 iNORTH + 7 iSOUTH + itu
where itPOP measures regional population, itDENS is the density of population in a
region, GV itApc reflects regional per capita income and2itGVApc the square of
regional per capita income. Finally, the dummy variables iCENTRAL , iNORTH and
iSOUTH are used in the model to catch a possible North-South distinction. The results
indicate that during the period 1977-1999, increases in regional growth tend to decrease
productive specialization initially but it rises at later stages of development. The same
is true for regional size as regional specialization falls with increases in regional
population. In addition, an important element of this study is the relation between
specialization and the geographical location of European regions. The findings reveal
that a possible centre-periphery gradient is evident in the model as Northern and
Southern regions present higher levels of regional specialization as compared to more
Central regions.
Regarding EU New Member-states, Kallioras and Petrakos (2010) test the industrial
growth performance17 for the regions of Hungary, Bulgaria, Romania, Estonia and
Slovenia during the early accession period, 1991-2000. The econometric model they
use takes the form:
kttrY _, =
n
trX1
,,
+ tr ,
17 The industrial growth of EU New Member-states is expressed in terms of employment data
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
30
where kttrY _, is the dependent industrial growth variable for region r and
n
trX1
,,
is the set of λ independent variables which are: Economic Integration18 with the
average EU-15 economy in the base year (1991), Regional Industrial Diversification19
(the inverse of regional specialization) in the base year, the share of Industrial
Employment in Capital-intensive Sectors in the total industrial employment, the
Average Size of Industrial Firms which accounts for possible economies of scale and
finally a Geographic Variable of the Relative Centrality of the EU NMS. From the
estimates, it seems that industrial diversification variable has a positive and statistically
significant effect on industrial employment growth. The authors try to interpret these
findings indicating that greater diversity in productive bases of NMS regions is better
for regional growth as it may act as a safeguard protecting the regions from possible
asymmetric shocks. From the rest of the variables only the Economic Integration
variable has a negative and statistically significant impact on regional employment
growth. The latter indicates that the exposure of weaker peripheral regions to new
market conditions has negatively affected them in terms of employment. Consequently,
there seems to be winners and losers from the process of European integration. Capital
regions and western regions that border the EU presented better growth potentials as
compared to the other more peripheral regions, mainly due to their favored geographic
location. Contrary to the previous study, Iara and Traistaru (2004) using regional data
for 20 NUTS III regions in Hungary over the period 1994-2000, find evidence of a
positive relationship between regional growth and regional manufacturing
specialization. However, the results in the last two surveys cannot be characterized as
comparable due to the fact that different dependent and explanatory variables are used
in the models, thereby changing the scope of each analysis.
2.4.4 The case of Greek regions
In this section, an overview of the available empirical literature with regard to Greek
regions – whose performance is the object of the dissertation – is presented. Greece,
18 Economic integration is expressed in the model with the use of an index of economicintegration (IEI), proposed by Petrakos et al. (2005)19 Regional industrial diversification is displayed with the use of Theil Entropy Index, proposedby Theil (1972)
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
31
which is considered to be the most peripheral country of the EU, could not sufficiently
deal with the new market conditions in the period after EU membership. The latter can
be attributed not only to its disadvantageous geographic position but also to the less
advanced industrial base in relation to the EU core and to the numerous structural
problems. It is evident that the share of industry in GDP has presented declining
trends20 throughout the period 1980-2000 and moreover, it is the lowest of all member
countries (Aiginger 2000). Despite the observed industrial decline, Greece has
exhibited an enormous speed of structural change, which nevertheless has no results in
terms of growth. This is probably due to the specialization of Greece in low growth
sectors, while the majority of member states follow high growth industries (Aiginger
2001). Indeed, Greek industrial structure seems to have been dominated by labour-
intensive sectors, as 50% of industrial GDP in 1985 has been concentrated in these
sectors (42% in only two sectors: Food, Beverages & Tobacco and Textiles & Wearing
Apparel) while the respective figure for the EU-15 is 36%. The overall image remains
almost the same in the year 2000, as labour-intensive sectors counts for the 45% of
industrial GDP in relation to 32% in EU-15 (Petrakos et al. 2005).
With respect to regional productive specialization for the period 1977-1999, Ezcurra et
al. (2004) point out that initially, Greek regions appeared to be more specialized in
comparison to the other European countries, but a tendency towards more
diversification and convergence with the European average took place during that
period. Using manufacturing employment data Brulhart (1998) finds that Greece has
presented the lowest specialization level in the European Union, thereby confirming the
view that a process of increasing diversification is evident from 1980 onwards. This
fact is also confirmed by Petrakos et al. (2006) who estimate regional diversification for
NUTS II and NUTS III Greek regions with the use of Theil index during the period
1980-2000. A closer look at the results reveals that the most urbanized regions (Athens,
Thessaloniki, Patra, Larissa and Volos) present more diversified structures as compared
to the other regions.
Finally, it will be very informative to present an econometric model of regional growth
performance in Greek regions proposed by Petrakos et al. (2005). They examine
20 From 14.59% in 1980 to 12.08% in 2000 (Petrakos et al. 2005)
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
32
manufacturing performance at NUTS III spatial level for the period 1981-2000 by using
as dependent variables the industrial GDP growth and labour productivity growth. The
explanatory variables used in the model are the following: regional diversification
expressed by Theil index, the average firm size of industrial firms, an index of
integration with the EU economy, an index of dissimilarity of regional structures in
comparison to EU economy, the shares in the tertiary sector, the shares in regional
productivity of the tertiary sector, the percentage of investment subsidized by the state
and per capita public investment by region. The results indicate that all variables –
except for index of integration and per capita public investment – have a statistically
significant and positive impact on regional growth. A more careful interpretation of the
results suggests that in the light of economic integration and fierce competition from
other European countries, increasing diversification and increasing dissimilarity to the
European average in combination with other factors was the key for better growth
performance. On the other side, regions which experienced increasing specialization
and similar industrial structures to the EU average faced with poor growth performance
and industrial decline. As a matter of fact, the results seem to confirm the view that a
more diversified production structure constitutes the appropriate solution for “weak”,
peripheral countries.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
33
CHAPTER 3: Data and methodology
The purpose of this section is to provide some useful explanations about the empirical
method and the data used in the analysis. It is basically attempted to be justified the
choice of an index which will be capable of explaining patterns of regional
specialization in Greek regions. A comprehensive presentation of the index and its
specific properties follows.
3.1 Indicators of regional specialization
A variety of indicators have been used in the literature in order to be determined the
spatial distribution of economic activity. A thorough analysis of the existing empirical
literature as regards regional specialization has been presented in the previous chapter
of literature review. The majority of surveys conducted include explanations of why
some indicators are better than others when patterns of regional specialization are
examined. A basic conclusion that can be securely inferred from these considerations is
that none of these measures can be regarded as optimal. Furthermore, very few attempts
have been undertaken to determine the criteria by which we should choose the
appropriate index21. Whatever the case may be, it is beyond any question that the
decision on which measure is the most appropriate for a specific survey depends highly
on the purpose of the investigation. Each measure presents specific properties, produces
different results and therefore may fit or may not fit to the purpose of a certain study.
For this purpose, a table which describes indicators that have been used most in the
existing empirical literature has been constructed (see Table 1 in the Appendix). The
table presents both absolute and relative measures of regional specialization describing
the mathematical form and the main characteristics of these indicators.
Obviously, the main distinction is between the so-called absolute and relative measures
of regional specialization22. Accordingly, the choice of the appropriate index constitutes
a trade-off procedure between absolute and relative measures. With respect to industrial
21 See Combes and Overman (2003) and Bode et al. (2004)22 The notions of absolute and relative specialization are described in the introduction of thepresent study.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
34
specialization, absolute indicators are based on the shares of individual industries
without taking into account a benchmark. This means that absolute measures do not
take into consideration the behavior of the broader geographical area (e.g. a country
when regions are under examination) and are based on shares which refer to a zero
distribution or a uniform distribution (Bode et al. 2004). A major advantage of absolute
indicators is that they measure the absolute size of specialization within a region, but on
the other side they do not allow for interregional comparisons of structural change. On
the other hand, relative indicators refer to the shares of individual industries according
to a reference distribution, and therefore they deal better with the internal of countries.
In this case relative specialization may be helpful if a comparison between different
regions in a country is attempted. Taking into account the above considerations, it can
be implied that absolute indicators focus on large countries as the degree of absolute
specialization will be proportional to the size of countries, while relative indicators give
more weight to small countries (Aiginger 1999).
3.2 Description of the methodology
Considering the merits of other indicators (see Table 1 in the Appendix) which have
been extensively used in the empirical literature, it is suggested that the most
appropriate index for the case of Greek regions is the Brülhart-Traeger-Theil index, the
general form of which is the following:
THEIL=
I
i i
i
i
ii
ara
ara
Nn ln
where I is the number of observations (the number of industries in the case of regional
specialization) investigated in the analysis, r is the region under examination, raiindicates the share of industry i in region r (in terms of employment) and ia denotes the
national share of industry i in the total manufacturing activity and functions as the
benchmark for the corresponding rai . In addition Nni represents the weighting factor
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
35
of Theil index such that i inN . The Nni ratio indicates the relative gravity each
industry presents (e.g. employment, production, area23) in relation to the total industrial
activity.
Theil index is characterized by substantial advantages, as compared to the other
measures of specialization and concentration. Simultaneously these advantages
constituted the basic criteria for the choice of Theil indicator. First of all, different types
of Theil indices can be estimated for different forms of specialization. This is to say that
Theil indicator can be measured for both types of specialization, absolute and relative
(Tsiapa 2008). Moreover, the relative indicator can be weighted by the share of each
industry in the total manufacturing providing in such a way another version of Theil
index. Secondly, a major advantage of Theil index not presented in other indicators is
the tendency to downgrade extreme observations due to its logarithmic form (Bode et
al. 2004). Another significant characteristic that all entropy measures24 deal with is the
ability of decomposition. According to its decomposition property, Theil index allows
for international, interregional and intertemporal comparisons (Bode et al. 2004). With
respect to regional specialization, decomposition property provides the ability of
estimation on both total spatial levels (e.g. comparison between regions) and
segmentary spatial levels (e.g. the internal of a region) [Tsiapa 2008]. Last but not least,
entropy measures present the capability to deal better with the Modifiable Area Unit
Problem known as MAUP in the literature. The use of entropy indices implies that they
may be estimated for different spatial levels (e.g. NUTS I, NUTS II or NUTS III spatial
level)25, but however this can lead to differentiated valuations and conclusions
regarding each spatial unit. Theil index partially reduces the intensity of this problem by
using as basic variable the number of employees or the area covered by each region
(Tsiapa 2008).
In our study, trends in regional specialization in Greek regions are estimated with the
use of the relative Theil index which takes the following form:
23 Area (square kilometers) cannot be used in the case of regional specialization as it is basicallyused for the estimation of spatial concentration (Topographic Theil index).24 Theil index belongs to the category of entropy measures.25 NUTS or Nomenclature of territorial units for statistics refers to the standard regionalclassification system used by Eurostat.
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
36
THEIL=
I
i i
ii a
rara ln
where ia r indicates the employment shares of industry i in the total manufacturing in
region r and ia refers to the employment shares of industry i in the total manufacturing
of Greek economy.
3.3 Data presentation
The main objective of this dissertation is to provide a comprehensive study of industrial
specialization patterns across Greek regions. For this purpose it has been proposed the
use of Theil entropy index as it is obviously more suitable than other conventional
measures to deal with the Greek case. The choice of Theil index was mainly due to its
desirable decomposition properties and its ability to downgrade the influences of
outliers. The estimation of Theil index is based on regional employment manufacturing
data for 51 NUTS III regions and 17 industrial sectors covering a period of 25 years,
from 1980 to 2005. With respect to the choice of the appropriate data set, employment
data are valuated as more preferable than other variables due to the fact that through the
use of employment, problems related to currency conversion and inflation rates – which
are inherent in value added and output data – can be avoided (Brülhart and Traeger
2003). Moreover, employment data can be characterized by “mobility”, an asset
inherent in employment which can provide a different viewpoint regarding the
inspection of industrial behavior.
It is worth noting that the period covered coincides with historical moments as regards
political and economic situation in Greece. The year 1980 constitutes a key point in
Greek history because one year later Greece joined officially the (then called) European
Economic Community. The enactment of the Single Market in 1992 and the entry of
Greece in the Economic and Monetary Union in 2001 also represent crucial points for
which manufacturing data are available. At this point, it should be pointed out that
trends in regional employment specialization are computed for the years 1980, 1985,
Institutional Repository - Library & Information Centre - University of Thessaly08/12/2017 19:04:07 EET - 137.108.70.7
-
37
1990, 1995, 2000 and 2005 catching in this way the possible effects EU agreements
could have on industrial structure of Greece before and after their implementation.
The dataset used in this study is from ELSTAT26 and consists of 17 manufacturing
branches (see Table 2 in the Appendix) following the Stakod 80 classification. It must
be referred that for the years 1980, 1985 and 1990 ELSTAT uses Stakod 80
classification, whereas for the years 1995, 2000 and 2005 ELSTAT uses Stakod 0327
classification. Originally Stakod 80 classification consists of 20 manufacturing
branches, however, they have been accumulated in 17 branches in order to be achieved
the best fit between Stakod 80 and Stakod 03. Although it is generally desirable “to
seek